Why video games launch unfinished

The economics behind the shift are not subtle. Adobe runs near 90% gross margin, Facebook 80%, Microsoft 70%, Google 60% — and Adobe now earns more from $9.99 a month than it did selling $700 Photoshop licences. Any industry watching those numbers was going to follow.

What travelled with the model was a culture. Ship fast, fail fast means products launch before they are ready and are fixed afterward, and anything that can be paywalled is — including features essential to using the service.

Games were a moonshot business before this, which is why publishers wanted out. Budgets ran $5M to $50M in the early 2010s, or $30M to $300M including marketing, spent over 2 to 5 years with no revenue until launch and no idea whether it would sell.

The development model was the safeguard, and it got replaced. Under waterfall you plan to a distant deadline and ship something stable; delays were common and acceptable. Under agile, work is broken into two-week sprints and shipping anything counts as a win — so launching incomplete became survivable rather than fatal.

CD Projekt Red is the control group: a studio that refused all of it. Single-player only, no microtransactions, no annual sequels, and a stated belief that a good game sells for years on its own.

It was right, spectacularly, once. Witcher 3 sold 6 million copies in six weeks where Witcher 2 took two years to sell 2.5 million, and 20 million by the end of 2015. Revenue went $30M to $177M — a 600% increase — net income $1M to $76M, and margin 5% to over 40%.

The returns justify the whole moonshot model. Witcher 3 cost $81 million and returned 90% in its first year, has passed 50 million copies, and is among the ten best-selling games ever. This is why Activision, EA and Ubisoft keep spending: you miss more than you hit, but one hit moves everything.

Then the same studio proved the risk. Cyberpunk 2077 took 8 years and $300 million and launched broken — and was still a commercial success at first, quadrupling revenue from $116M to $486M and taking net income from $40M to $255M.

The bill came the year after: revenue and game sales fell over 60% in 2021 — double the decline Witcher 3 saw at the same point. CD Projekt Red then cancelled Cyberpunk's live service entirely and committed to agile. The trust, not the launch, was the loss.

The structural problem is the schedule. Shipping two games a decade means you have to get at least one right; get both wrong and the studio is finished. Every extra year in development raises the stakes on an already binary bet.

Square Enix diagnosed all of this correctly, in public, in 2013 — after a $100 million loss — naming "an inflexible earnings model and long-term, large-scale development resulting in a low rate of investment turnover" and calling for a move off disc-based earnings.

Its 2016 blockbuster slate looked like vindication: game sales up 42% from $860M to $1.2B, net income doubling $75M to $153M. Behind the numbers the lineup was failing. Hitman shipped episodically, sold badly, and Square Enix sold the studio and franchise and wrote down $43 million — later admitting it needed four more years of sales just to break even.

Deus Ex: Mankind Divided is the clearest case of monetization applied to a game that could not carry it: microtransactions in a single-player campaign plus a free-to-play mode where players could not even see each other, forced on the studio at short notice. 5.5 years, $55 million, needed 3 million copies to break even, sold 2 million.

So Square Enix went all-in, handing its best studios three years and a blank cheque for a Marvel live service. Marvel's Avengers cost $170-190 million, shed 96% of its Steam players within two months — down to about 1,000 concurrent — sold 3 million copies, and produced a $105 million write-down. Going all-in on SaaS is not the same as being good at it.

Take-Two is the counterexample, and it started early and small. It tested season passes and virtual currency on middle-tier franchises — Borderlands 2 with 16 DLC releases, NBA 2K13 with the franchise's first virtual currency — before applying anything to its crown jewels.

The result is a company that earns more each year from games released a decade ago. Recurrent consumer spending grew about 60% a year from 2014 to 2017, going from 11% of revenue to nearly half, then from $1.3B to $2.1B in 2021 to become 62% of total revenue — the first time it beat full game sales. GTA V made $1 billion in three days on a $265M budget and $6 billion by 2018; the difference from Square Enix is that Take-Two built live services into games that could hold them, rather than bolting them onto games that could not.